【摘要】 针对使用惯性测量单元（IMU）的计步算法在多行人条件下存在的伪零速造成计步准确率低的问题，提出了一种后向窗口滤波优化计步方法。首先，利用巴特沃斯低通滤波对IMU输出的加速度和角速度数据进行平滑处理，计算合加速度模值与合角速度模值。然后，采用阈值和零速检测算法得到初始步数识别数组。最后，通过计算后向窗口内的步数均值，取整优化步数识别数组，抑制零速检测中伪零速的干扰。在六名实验者的足部安装IMU模块，在单行人或多行人条件下以连续的多种行走活动进行实验验证。实验结果表明，所提方法的计步平均准确率为99.8%，相比多源信息自适应步数检测方法，准确率提高了28.5%，具有较高的计步准确率。更多还原

【Abstract】 In response to the problem of low accuracy of step counting algorithm using inertial measurement unit(IMU) under the condition of multiple pedestrians caused by false zero, a backward window filtering optimization step counting method is proposed. Firstly, Butterworth low-pass filtering is used to smooth the three axis acceleration and angular velocity output by the IMU, and the resultant acceleration modulus and resultant angular velocity modulus are calculated separately. Then, the initial step recognition array is obtained using threshold and zero velocity detection algorithms. Finally, by calculating the average number of steps in the backward window, the step recognition array is rounded and optimized to eliminate the interference problem of false zero in zero velocity detection. The IMU is installed on the feet of six participants, and the experimental verification is carried out with continuous multiple walking activities under single or multipedestrian conditions. The experimental results show that the proposed method has an average step counting accuracy of 99.8%. Compared with the multi-source information adaptive step detection method, the accuracy of the proposed method is improved by 28.5%, which has high step counting accuracy.更多还原